Prashik Mendhe

-

Enhanced framework to meet industry benchmarks.

Results-driven Credit Risk Modelling professional with 3+ years of experience leading model development, migration, and enhancement for global portfolios. Demonstrated expertise in delivering quantifiable improvements in model performance, data coverage, and business value. Adept at cross-functional leadership, technical innovation, and translating analytics into actionable strategy. Skilled in Python, SQL, SAS, and advanced statistical modelling.

-

## Key Achievements

- **Delivered $2M PTI benefit** by creating and deploying the Modified Revolve Index for US and International Markets, ranking among the top 10 predictive variables.

- **Doubled Fundamental Rating variable coverage** (US: 2,624→4,821; Intl: 1,738→3,109) by identifying and resolving logic gaps in GCP models and integrating ICRUSE and GRU data.

- **Boosted model GINI by 10 percentage points** and fill rates from 73% to 92% through engineering and deploying new tenure and paydown variables.

- **Elevated high-risk segment capture rates by 1.43%** by integrating new predictive variables into US small business models.

- **Sized and optimized real-time vs batch attribute performance**, minimizing loss exposure during platform migration.

- **Resolved critical paydown issues** during AMP migration through targeted case reviews, ensuring accurate risk assessment and seamless deployment.

- **Designed next-generation closed-loop variables** for AMP, leveraging unpaid period models to enhance predictive accuracy.

- **Pioneered new approaches to model performance analysis** by developing transaction-level methodologies for pend/deny and utilization variables, introducing the "Early Capture" metric and uncovering nuanced impacts on risk identification and model sensitivity.

- **Maintained 100% attendance and upheld highest standards of punctuality and ethics**, supporting a culture of reliability and accountability.

-

## Technical Skills

- **Programming/Tools:** Python, SQL, SAS, R, Spark, Excel, Tableau

- **Statistical Methods:** Logistic Regression, Scorecard Development, Feature Engineering, GINI, KS, Reject Inference

- **Platforms:** GCP, AWS, ODL, CAS, AMP

- **Other:** UAT, PIV, Model Validation, Data Pipeline Engineering

-

## Professional Experience

**Credit Risk Modelling Analyst**

American Express | [City, State] | Jul 2022 – Present

**Model Innovation & Impact**

- Spearheaded the migration of Prob-R, CDSS, and TSR models to the AMP platform, impacting 10M+ customer accounts and ensuring robust model performance for POS, Credit Concern, and CIUMC journeys.

- Engineered and deployed Best Tenure, Hcurtd01, and Hcurvindx variables, increasing fill rates from 73% to 92% and GINI from 23% to 32% in international SBS models.

- Identified and piloted new variables for high-balance, low-tenure segments, enhancing early risk detection and model accuracy.

- Sized real-time vs batch attributes to quantify and minimize loss during migration to the new platform.

- Designed and implemented next-generation closed-loop variables for AMP, improving predictive power by leveraging unpaid period models.

- Created and institutionalized the Modified Revolve Index, delivering $2M PTI benefit and ranking among top 10 predictive variables for Open portfolios.

**Advanced Model Performance Analysis (Pend/Deny & Utilization)**

- Developed and executed a novel transaction-level analysis approach for pend/deny and utilization variables, moving beyond traditional random-sample methods to assess model performance with greater precision.

- Introduced the "Early Capture" metric to measure the percentage of lending defaulters identified with and without pend/deny models, revealing a 2.3% relative capture lift for customers with zero Ac99dpdy and a 1.96% drop for non-zero cases—insights that directly informed model calibration and business strategy.

- Led sensitivity studies on the Utilization variable, showing that while it boosts scores for high-utilization customers, its influence diminishes at lower utilization, and overall model capture differences remain marginal (2% vs 3%).

- Recommended differentiated model strategies: deploying CDSS models with Utilization for transaction approvals in high-utilization bands, and TSR models without Utilization for structural decisions like line reduction in low-utilization bands.

**Leadership & Collaboration**

- Led cross-functional teams (Open, SBS, GCP, Data Stewards) in end-to-end model migration and variable deployment, ensuring consensus on logic changes and seamless integration.

- Facilitated UAT and PIV for all migrated variables, achieving zero post-implementation defects and full stakeholder buy-in.

- Mentored junior analysts and led regular case review sessions, fostering a culture of continuous improvement and innovation.

- Advocated for and implemented monthly SBFE triggers using DNB, Experian, and Equifax data, increasing trigger effectiveness and reducing daily trigger dependency by 30%.

**Process Optimization & Ethics**

- Streamlined model attribute lists, removing redundancies and increasing efficiency across global platforms.

- Maintained an outstanding attendance record and consistently modeled punctuality and ethical standards, reinforcing team reliability.